diff --git a/jsonpickle/ext/pandas.py b/jsonpickle/ext/pandas.py
index 18f43a57..9154f745 100644
--- a/jsonpickle/ext/pandas.py
+++ b/jsonpickle/ext/pandas.py
@@ -56,7 +56,7 @@ def restore_pandas(self, data):
 def make_read_csv_params(meta):
     meta_dtypes = meta.get('dtypes', {})
     # The header is used to select the rows of the csv from which
-    # the columns names are retrived
+    # the columns names are retrieved
     header = meta.get('header', [0])
     parse_dates = []
     converters = {}
@@ -102,7 +102,9 @@ def flatten(self, obj, data):
     def restore(self, data):
         csv, meta = self.pp.restore_pandas(data)
         params, timedeltas = make_read_csv_params(meta)
-        column_levels_names = meta.get('column_level_names', None)
+        # None makes it compatible with objects serialized before
+        # column_levels_names has been introduced.
+        column_level_names = meta.get('column_level_names', None)
         df = (
             pd.read_csv(StringIO(csv), **params)
             if data['values'].strip()
@@ -113,7 +115,8 @@ def restore(self, data):
 
         df.set_index(decode(meta['index']), inplace=True)
         # restore the column level(s) name(s)
-        df.columns.names = column_levels_names
+        if column_level_names:
+            df.columns.names = column_level_names
         return df
 
 
diff --git a/tests/pandas_test.py b/tests/pandas_test.py
index 0cf4cce6..b0258cfc 100644
--- a/tests/pandas_test.py
+++ b/tests/pandas_test.py
@@ -15,6 +15,7 @@
 
 import jsonpickle
 import jsonpickle.ext.pandas
+from jsonpickle.compat import PY2
 
 
 @pytest.fixture(scope='module', autouse=True)
@@ -289,5 +290,24 @@ def test_dataframe_with_timedelta64_dtype():
     assert data_frame['Duration'][2] == actual['Duration'][2]
 
 
+def test_multilevel_columns():
+    if not PY2:
+        pytest.skip('This test does not yet pass on Python 3')
+
+    iterables = [['inj', 'prod'], ['hourly', 'cumulative']]
+    names = ['first', 'second']
+    # transform it to tuples
+    columns = pd.MultiIndex.from_product(iterables, names=names)
+    # build a multi-index from it
+    data_frame = pd.DataFrame(
+        np.random.randn(3, 4), index=['A', 'B', 'C'], columns=columns
+    )
+    encoded = jsonpickle.encode(data_frame)
+    cloned_data_frame = jsonpickle.decode(encoded)
+    assert isinstance(cloned_data_frame, pd.DataFrame)
+    assert data_frame.columns.names == cloned_data_frame.columns.names
+    assert_frame_equal(data_frame, cloned_data_frame)
+
+
 if __name__ == '__main__':
     pytest.main([__file__])